1. Field of the Invention
The present invention relates to refresh technique for a semiconductor memory device such as a DRAM (Dynamic Random Access Memory), particularly, a semiconductor memory device having a configuration to perform both an auto refresh in normal operation and a self refresh in standby state, and refresh control method thereof.
2. Description of the Related Art
Recently, reduction in consumption current in standby state is strongly required for a large capacity DRAM mounted on a mobile device. Thus, as an effective scheme for achieving lower power consumption of the DRAM for the mobile devices, partial array self refresh function (hereinafter referred to as “PASR function”) is known. This PASR function performs a selective self refresh in which, in a memory cell array which is divided into a plurality of banks, one or some of the banks of the memory cell array are selectively refreshed. In this case, refresh target for retaining data is limited to one or some of the banks, long period refresh operation is performed only for the corresponding banks sequentially. For example, two banks among four banks are designated to be refreshed, refresh area is reduced by half relative to normal refresh. Thus, the employment of the PASR function is effective to reduce consumption current in the DRAM. Various refresh control techniques including the PASR function for selectively performing refresh for a partial area of the memory cell array have been proposed (see Japanese Patent Laid-Open No. Hei 2-192096, Japanese Patent Laid-Open No. 2002-334576, and Japanese Patent Laid-Open No 2004-259343).
However, when the PASR function is employed in the conventional DRAM, a specification is generally used for the purpose of an application in the self refresh, in which banks to be refreshed cannot be designated in the auto refresh in normal operation. In the auto refresh in normal operation, requirement for employing the PASR function in terms of reducing consumption current is not very large compared with the self refresh. However, a use form for limiting a refresh target to part of the banks is assumed for a reason concerning system or process. For example, if a situation arises in which a specific portion of the memory cell array is accessed within a certain time using image data and the like in such a use form, the auto refresh operation is not performed for this specific portion in normal operation, and thereby reducing the number of refresh operations as a whole so as to improve the use efficiency. However, even when the designation of banks based on the PASR function is applied to the auto refresh, portions requiring refresh and other portions not requiring refresh generally do not conform to each other. Thus, it is difficult to achieve an optimum refresh control corresponding to the respective situations of the self refresh and the auto refresh.
Further, although whether or not the refresh operation is performed is set for each bank in the above-mentioned PASR function, a case is assumed in which actually portions requiring refresh and other portions not requiring refresh are actually mixed with each other. In such a case, for example, image data or the like is distributed and stored in respective banks for accessing in a bank interleave mode. If whether or not the refresh operation is performed is set for each bank in this case, it is a problem that the number of banks which can be used decreases and the access efficiency is reduced.
Furthermore, in the conventional PASR function described above, portions to be refreshed can be limited, but a refresh period of each memory cell included in the limited portions is not considered to be selectively set. That is, when the entire number of refresh operations is reduced by limiting the portions to be refreshed, it is not possible to select either of two cases, in one of which the refresh period is maintained while the access efficiency is improved by reducing a refresh busy rate as a whole, and in the other of which the refresh busy rate is maintained while operating margin is improved by shortening the refresh period for the memory cells. Particularly, in the auto refresh, it is required to apply each of the two cases according to the use form. However, it is a problem that the conventional PASR function described above cannot achieve this requirement.